Fix non-deterministic collection order for module-scanned injectables

_find_objects_in_module collected discovered classes into a set, so
Sequence[T]/Mapping[T] ordering was hash-order (id()-derived) rather than
the documented registration order whenever injectables came from a
scanned module instead of an explicit list. Directory walks also relied
on Path.iterdir(), which isn't guaranteed stable.

Switch discovery to an insertion-ordered dict-as-set and sort the
directory walk, so scanned modules produce deterministic results.

Fixes #144

Author: afonsojanu

test/unit/services/collection_scan/__init__.py

@@ -0,0 +1,31 @@
+from typing import Protocol
+
+from wireup import injectable
+
+
+class Greeter(Protocol):
+    def hi(self) -> str: ...
+
+
+@injectable(as_type=Greeter, qualifier="delta")
+class Delta:
+    def hi(self) -> str:
+        return "delta"
+
+
+@injectable(as_type=Greeter, qualifier="beta")
+class Beta:
+    def hi(self) -> str:
+        return "beta"
+
+
+@injectable(as_type=Greeter)
+class Alpha:
+    def hi(self) -> str:
+        return "alpha"
+
+
+@injectable(as_type=Greeter, qualifier="gamma")
+class Gamma:
+    def hi(self) -> str:
+        return "gamma"

test/unit/services/collection_scan/greeters.py

@@ -0,0 +1,20 @@
+import unittest
+from collections.abc import Sequence
+
+import wireup
+
+from test.unit.services.collection_scan import greeters
+
+
+class DiscoveryOrderTest(unittest.TestCase):
+    def test_module_scan_collection_injection_has_deterministic_order(self):
+        container = wireup.create_sync_container(injectables=[greeters])
+        result = [g.hi() for g in container.get(Sequence[greeters.Greeter])]
+
+        self.assertEqual(["alpha", "beta", "delta", "gamma"], result)
+
+    def test_module_scan_discovery_order_is_stable_across_repeated_scans(self):
+        first = [g.hi() for g in wireup.create_sync_container(injectables=[greeters]).get(Sequence[greeters.Greeter])]
+        second = [g.hi() for g in wireup.create_sync_container(injectables=[greeters]).get(Sequence[greeters.Greeter])]
+
+        self.assertEqual(first, second)

test/unit/test_discovery_order.py

@@ -24,9 +24,9 @@ def _is_valid_wireup_target(obj: Any) -> bool:
         # "from flask import g" would cause a hasattr call to g outside of app context.
         return (isinstance(obj, FunctionType) or inspect.isclass(obj)) and hasattr(obj, "__wireup_registration__")
 
-    all_targets = {
-        m for module in injectable_modules for m in _find_objects_in_module(module, predicate=_is_valid_wireup_target)
-    }
+    all_targets: dict[type, None] = {}
+    for module in injectable_modules:
+        all_targets.update(dict.fromkeys(_find_objects_in_module(module, predicate=_is_valid_wireup_target)))
 
     for cls in all_targets:
         reg = getattr(cls, "__wireup_registration__", None)
@@ -39,14 +39,14 @@ def _is_valid_wireup_target(obj: Any) -> bool:
     return abstract_registrations, injectable_registrations
 
 
-def _find_objects_in_module(module: ModuleType, predicate: Callable[[Any], bool]) -> set[type]:
-    classes: set[type[Any]] = set()
+def _find_objects_in_module(module: ModuleType, predicate: Callable[[Any], bool]) -> list[type]:
+    classes: dict[type[Any], None] = {}
 
-    def _module_get_objects(m: ModuleType) -> set[type]:
-        return {obj for _, obj in inspect.getmembers(m) if predicate(obj)}
+    def _module_get_objects(m: ModuleType) -> list[type]:
+        return [obj for _, obj in inspect.getmembers(m) if predicate(obj)]
 
     def _find_in_path(path: Path, parent_module_name: str) -> None:
-        for file in path.iterdir():
+        for file in sorted(path.iterdir()):
             if file.name == "__pycache__":
                 continue
 
@@ -60,12 +60,12 @@ def _find_in_path(path: Path, parent_module_name: str) -> None:
                     parent_module_name if file.name == "__init__.py" else f"{parent_module_name}.{file.name[:-3]}"
                 )
                 sub_module = importlib.import_module(full_module_name)
-                classes.update(_module_get_objects(sub_module))
+                classes.update(dict.fromkeys(_module_get_objects(sub_module)))
 
     if f := module.__file__:
         if f.endswith("__init__.py"):
             _find_in_path(Path(f).parent, module.__name__)
         else:
-            classes.update(_module_get_objects(module))
+            classes.update(dict.fromkeys(_module_get_objects(module)))
 
-    return classes
+    return list(classes)